NetworkAnalysis

Class for doing network analysis.

class NetworkAnalysis(network, rules, log=True, detailed_log=False)[source]

Bases: object

Class for doing network analysis.

Parameters:
  • network (GeoDataFrame) – A GeoDataFrame of line geometries.

  • rules (NetworkAnalysisRules | dict) – The rules for the analysis, either as an instance of NetworkAnalysisRules or a dictionary with the parameters as keys.

  • log (bool) – If True (default), a DataFrame with information about each analysis run will be stored in the ‘log’ attribute.

  • detailed_log (bool) – If True, the log DataFrame will include columns for all arguments passed to the analysis method, plus standard deviation and percentiles (25th, 50th, 75th) of the weight column in the results. Defaults to False.

The class implements methods for doing network analysis based on GeoDataFrames of origin and destination points.

The ‘od_cost_matrix’ method is the fastest, and returns a DataFrame with only indices and travel costs between each origin-destination pair.

The ‘get_route’ method does the same, but also returns the line geometry of the routes. ‘get_k_routes’ can be used to find multiple routes between each OD pair.

The service_area methods only take a set of origins, and return the lines that can be reached within one or more breaks.

The ‘get_route_frequencies’ method is a bit different. It returns the individual line segments that were visited with an added column for how many times the segments were used.

network

A Network instance that holds the lines and nodes (points).

rules

NetworkAnalysisRules instance.

log

A DataFrame with information about each analysis run.

Examples:

Read example data.

>>> import sgis as sg
>>> roads = sg.read_parquet_url("https://media.githubusercontent.com/media/statisticsnorway/ssb-sgis/main/tests/testdata/roads_oslo_2022.parquet")

Preparing the lines for directed network analysis.

>>> connected_roads = sg.get_connected_components(roads).query("connected == 1")
>>> directed_roads = sg.make_directed_network(
...     connected_roads,
...     direction_col="oneway",
...     direction_vals_bft=("B", "FT", "TF"),
...     minute_cols=("drivetime_fw", "drivetime_bw"),
...     dropnegative=True,
...     dropna=True,
... )
>>> rules = sg.NetworkAnalysisRules(weight="minutes", directed=True)
>>> nwa = sg.NetworkAnalysis(network=directed_roads, rules=rules, detailed_log=False)
>>> nwa
NetworkAnalysis(
    network=Network(6364 km, percent_bidirectional=87),
    rules=NetworkAnalysisRules(weight=minutes, directed=True, search_tolerance=250, search_factor=0, split_lines=True, ...),
    log=True, detailed_log=False,
)

Now we’re ready for network analysis.

copy(deep=True)[source]

Returns a (deep) copy of the class instance.

Parameters:

deep (bool) – Whether to return a deep or shallow copy. Defaults to True.

Return type:

NetworkAnalysis

get_k_routes(origins, destinations, *, k, drop_middle_percent, rowwise=False, destination_count=None, cutoff=None)[source]

Returns the geometry of 1 or more routes between origins and destinations.

Finds the route with the lowest cost (minutes, meters, etc.) from a set of origins to a set of destinations. Then the middle part of the route is removed from the graph the new low-cost path is found. Repeats k times. If k=1, it is identical to the get_route method.

Parameters:
  • origins (GeoDataFrame) – GeoDataFrame of points from where the routes will originate.

  • destinations (GeoDataFrame) – GeoDataFrame of points from where the routes will terminate.

  • k (int) – the number of low-cost routes to find.

  • drop_middle_percent (int) – how many percent of the middle part of the routes that should be removed from the graph before the next k route is calculated. If set to 100, only the median edge will be removed. If set to 0, all but the first and last edge will be removed. The graph is copied for each od pair.

  • rowwise (bool) – if False (default), it will calculate the cost from each origins to each destination. If true, it will calculate the cost from origin 1 to destination 1, origin 2 to destination 2 and so on.

  • destination_count (int | None) – number of closest destinations to keep for each origin. If None (default), all trips will be included. The number of destinations might be higher than the destination_count if trips have equal cost.

  • cutoff (int | float | None) – the maximum cost (weight) for the trips. Defaults to None, meaning all rows will be included. NaNs will also be removed if cutoff is specified.

Return type:

GeoDataFrame

Returns:

A DataFrame with the geometry of the k routes between origin and destination. Also returns the column ‘k’, a weight column and the columns ‘origin’ and ‘destination’, containing the indices of the origins and destinations GeoDataFrames.

Note

How many percent of the route to drop from the graph, will determine how many k routes will be found. If 100 percent of the route is dropped, it is very hard to find more than one path for each OD pair. If ‘drop_middle_percent’ is 1, the resulting routes might be very similar, depending on the layout of the network.

Raises:

ValueError – if drop_middle_percent is not between 0 and 100.

Parameters:
  • origins (GeoDataFrame)

  • destinations (GeoDataFrame)

  • k (int)

  • drop_middle_percent (int)

  • rowwise (bool)

  • destination_count (int | None)

  • cutoff (int | float | None)

Return type:

GeoDataFrame

Examples:

Create the NetworkAnalysis instance.

>>> import sgis as sg
>>> roads = sg.read_parquet_url('https://media.githubusercontent.com/media/statisticsnorway/ssb-sgis/main/tests/testdata/roads_oslo_2022.parquet')
>>> directed_roads = sg.get_connected_components(roads).loc[lambda x: x["connected"] == 1].pipe(sg.make_directed_network_norway, dropnegative=True)
>>> rules = sg.NetworkAnalysisRules(weight="minutes", directed=True)
>>> nwa = sg.NetworkAnalysis(network=directed_roads, rules=rules, detailed_log=False)

Getting 10 fastest routes from one point to another point.

>>> points = sg.read_parquet_url('https://media.githubusercontent.com/media/statisticsnorway/ssb-sgis/main/tests/testdata/points_oslo.parquet')
>>> point1 = points.iloc[[0]]
>>> point2 = points.iloc[[1]]
>>> k_routes = nwa.get_k_routes(
...             point1,
...             point2,
...             k=10,
...             drop_middle_percent=1
... )
>>> print(k_routes.to_string(max_rows=10, max_colwidth=50, show_dimensions=True))
   origin  destination    minutes   k                                           geometry
0       0            1  13.209125   1  MULTILINESTRING Z ((263167.581 6651182.119 NaN...
1       0            1  14.253618   2  MULTILINESTRING Z ((263167.581 6651182.119 NaN...
2       0            1  14.407402   3  MULTILINESTRING Z ((263167.581 6651182.119 NaN...
3       0            1  15.066977   4  MULTILINESTRING Z ((263167.581 6651182.119 NaN...
4       0            1  15.131888   5  MULTILINESTRING Z ((263167.581 6651182.119 NaN...
5       0            1  15.593228   6  MULTILINESTRING Z ((263167.581 6651182.119 NaN...
6       0            1  16.386565   7  MULTILINESTRING Z ((263167.581 6651182.119 NaN...
7       0            1  16.653277   8  MULTILINESTRING Z ((263167.581 6651182.119 NaN...
8       0            1  16.682548   9  MULTILINESTRING Z ((263167.581 6651182.119 NaN...
9       0            1  17.956501  10  MULTILINESTRING Z ((263167.581 6651182.119 NaN...

[10 rows x 5 columns]

We got all 10 routes because only the middle 1 percent of the routes are removed in each iteration. Let’s compare with dropping middle 50 and middle 100 percent.

>>> k_routes = nwa.get_k_routes(
...             point1,
...             point2,
...             k=10,
...             drop_middle_percent=50
...         )
>>> print(k_routes.to_string(max_rows=10, max_colwidth=50, show_dimensions=True))
   origin  destination    minutes  k                                           geometry
0       0            1  13.209125  1  MULTILINESTRING Z ((263167.581 6651182.119 NaN...
1       0            1  14.445367  2  MULTILINESTRING Z ((263167.581 6651182.119 NaN...
2       0            1  20.308589  3  MULTILINESTRING Z ((263167.581 6651182.119 NaN...
3       0            1  23.676072  4  MULTILINESTRING Z ((263167.581 6651182.119 NaN...

[4 rows x 5 columns]
>>> k_routes = nwa.get_k_routes(
...             point1,
...             point2,
...             k=10,
...             drop_middle_percent=100
...         )
>>> print(k_routes.to_string(max_rows=10, max_colwidth=50, show_dimensions=True))
   origin  destination    minutes  k                                           geometry
0       0            1  13.209125  1  MULTILINESTRING Z ((263167.581 6651182.119 NaN...

[1 rows x 5 columns]
get_route(origins, destinations, *, rowwise=False, destination_count=None, cutoff=None, n_jobs=None)[source]

Returns the geometry of the low-cost route between origins and destinations.

Finds the route with the lowest cost (minutes, meters, etc.) from a set of origins to a set of destinations. If the weight is meters, the shortest route will be found. If the weight is minutes, the fastest route will be found.

Parameters:
  • origins (GeoDataFrame) – GeoDataFrame of points from where the routes will originate

  • destinations (GeoDataFrame) – GeoDataFrame of points from where the routes will terminate.

  • rowwise (bool) – if False (default), it will calculate the cost from each origins to each destination. If true, it will calculate the cost from origin 1 to destination 1, origin 2 to destination 2 and so on.

  • destination_count (int | None) – number of closest destinations to keep for each origin. If None (default), all trips will be included. The number of destinations might be higher than the destination_count if trips have equal cost.

  • cutoff (int | float | None) – the maximum cost (weight) for the trips. Defaults to None, meaning all rows will be included. NaNs will also be removed if cutoff is specified.

  • n_jobs (int | None) – Number of parallell jobs.

Return type:

GeoDataFrame

Returns:

A DataFrame with the geometry of the routes between origin and destination. Also returns a weight column and the columns ‘origin’ and ‘destination’, containing the indices of the origins and destinations GeoDataFrames.

Examples:

Create the NetworkAnalysis instance.

>>> import sgis as sg
>>> roads = sg.read_parquet_url("https://media.githubusercontent.com/media/statisticsnorway/ssb-sgis/main/tests/testdata/roads_oslo_2022.parquet")
>>> directed_roads = sg.get_connected_components(roads).loc[lambda x: x["connected"] == 1].pipe(sg.make_directed_network_norway, dropnegative=True)
>>> rules = sg.NetworkAnalysisRules(weight="minutes", directed=True)
>>> nwa = sg.NetworkAnalysis(network=directed_roads, rules=rules, detailed_log=False)

Get routes from one origin to ten destination points.

>>> points = sg.read_parquet_url("https://media.githubusercontent.com/media/statisticsnorway/ssb-sgis/main/tests/testdata/points_oslo.parquet")
>>> routes = nwa.get_route(points.iloc[[0]], points.iloc[:10])
>>> print(routes.to_string(max_rows=10, max_colwidth=50, show_dimensions=True))
   origin  destination    minutes                                           geometry
0       0            1  13.209125  MULTILINESTRING ((263167.581 6651182.119, 2631...
1       0            2  10.994973  MULTILINESTRING ((263167.581 6651182.119, 2631...
2       0            3   7.749668  MULTILINESTRING ((263167.581 6651182.119, 2631...
3       0            4  14.849464  MULTILINESTRING ((263167.581 6651182.119, 2631...
4       0            5  14.391621  MULTILINESTRING ((263167.581 6651182.119, 2631...
5       0            6  13.071382  MULTILINESTRING ((263167.581 6651182.119, 2631...
6       0            7   4.439742  MULTILINESTRING ((263167.581 6651182.119, 2631...
7       0            8  13.673451  MULTILINESTRING ((263167.581 6651182.119, 2631...
8       0            9   8.275748  MULTILINESTRING ((263167.581 6651182.119, 2631...

[9 rows x 4 columns]
get_route_frequencies(origins, destinations, weight_df=None, default_weight=None, rowwise=False, strict=False, frequency_col='frequency', n_jobs=None)[source]

Finds the number of times each line segment was visited in all trips.

Finds the route with the lowest cost (minutes, meters, etc.) from a set of origins to a set of destinations and summarises the number of times each segment was used. The aggregation is done on the line indices, which is much faster than getting the geometries and then dissolving.

The trip frequencies can be weighted (multiplied) based on ‘weight_df’. See example below.

Parameters:
  • origins (GeoDataFrame) – GeoDataFrame of points from where the routes will originate.

  • destinations (GeoDataFrame) – GeoDataFrame of points from where the routes will terminate.

  • weight_df (DataFrame | None) – A long formated DataFrame where each row contains the indices of an origin-destination pair and the number to multiply the frequency for this route by. The DataFrame can either contain three columns (origin index, destination index and weight. In that order) or only a weight column and a MultiIndex where level 0 is origin index and level 1 is destination index.

  • default_weight (int | float | None) – If set, OD pairs not represented in ‘weight_df’ will be given a default weight value.

  • rowwise (bool) – if False (default), it will calculate the cost from each origins to each destination. If true, it will calculate the cost from origin 1 to destination 1, origin 2 to destination 2 and so on.

  • strict (bool) – If True, all OD pairs must be in weigth_df if specified. Defaults to False.

  • frequency_col (str) – Name of column with the number of times each road was visited. Defaults to ‘frequency’.

  • n_jobs (int | None) – Number of parallell jobs.

Return type:

GeoDataFrame

Returns:

A GeoDataFrame with all line segments that were visited at least once, with a column with the number of times the line segment was used in the individual routes.

Note

The resulting lines will keep all columns of the ‘gdf’ of the Network.

Raises:

ValueError – If weight_df is not a DataFrame with one or three columns that contain weights and all indices of ‘origins’ and ‘destinations’.

Parameters:
  • origins (GeoDataFrame)

  • destinations (GeoDataFrame)

  • weight_df (DataFrame | None)

  • default_weight (int | float | None)

  • rowwise (bool)

  • strict (bool)

  • frequency_col (str)

  • n_jobs (int | None)

Return type:

GeoDataFrame

Examples:

Create the NetworkAnalysis instance.

>>> import sgis as sg
>>> import pandas as pd
>>> roads = sg.read_parquet_url("https://media.githubusercontent.com/media/statisticsnorway/ssb-sgis/main/tests/testdata/roads_oslo_2022.parquet")
>>> directed_roads = sg.get_connected_components(roads).loc[lambda x: x["connected"] == 1].pipe(sg.make_directed_network_norway, dropnegative=True)
>>> rules = sg.NetworkAnalysisRules(weight="minutes", directed=True)
>>> nwa = sg.NetworkAnalysis(network=directed_roads, rules=rules, detailed_log=False)

Get some points.

>>> points = sg.read_parquet_url("https://media.githubusercontent.com/media/statisticsnorway/ssb-sgis/main/tests/testdata/points_oslo.parquet")
>>> origins = points.iloc[:3]
>>> destinations = points.iloc[3:6]

Get number of times each road was visited for trips from three to three points. Count the roads for each frequency, independently of the ordering of roads with equal frequencies.

>>> frequencies = nwa.get_route_frequencies(origins, destinations)
>>> frequencies.groupby("frequency").size().to_dict()
{1.0: 277, 2.0: 425, 3.0: 173, 4.0: 112}
>>> unweighted = frequencies["frequency"].sort_index()

The frequencies can be weighted for each origin-destination pair by specifying ‘weight_df’. This can be a DataFrame with three columns, where the first two contain the indices of the origin and destination (in that order), and the third the number to multiply the frequency by. ‘weight_df’ can also be a DataFrame with a 2-leveled MultiIndex, where level 0 is the origin index and level 1 is the destination.

Constructing a DataFrame with all od-pair combinations and give all rows a weight of 10.

>>> od_pairs = pd.MultiIndex.from_product(
...     [origins.index, destinations.index], names=["origin", "destination"]
... )
>>> weight_df = pd.DataFrame(index=od_pairs).reset_index()
>>> weight_df["weight"] = 10
>>> print(weight_df.to_string(max_rows=10, max_colwidth=50, show_dimensions=True))
   origin  destination  weight
0       0            3      10
1       0            4      10
2       0            5      10
3       1            3      10
4       1            4      10
5       1            5      10
6       2            3      10
7       2            4      10
8       2            5      10

[9 rows x 3 columns]

All frequencies will now be multiplied by 10.

>>> frequencies = nwa.get_route_frequencies(origins, destinations, weight_df=weight_df)
>>> frequencies.groupby("frequency").size().to_dict()
{10.0: 277, 20.0: 425, 30.0: 173, 40.0: 112}
>>> pd.testing.assert_series_equal(frequencies["frequency"].sort_index(), unweighted * 10)

‘weight_df’ can also be a DataFrame with one column (the weight) and a MultiIndex.

>>> weight_df = pd.DataFrame(index=od_pairs)
>>> weight_df["weight"] = 10
>>> weight_df
                    weight
origin destination
0      3                10
       4                10
       5                10
1      3                10
       4                10
       5                10
2      3                10
       4                10
       5                10
od_cost_matrix(origins, destinations, *, rowwise=False, destination_count=None, cutoff=None, lines=False)[source]

Fast calculation of many-to-many travel costs.

Finds the the lowest cost (minutes, meters, etc.) from a set of origins to a set of destinations. The index of the origins and destinations are used as values for the returned columns ‘origins’ and ‘destinations’.

Parameters:
  • origins (GeoDataFrame) – GeoDataFrame of points from where the trips will originate

  • destinations (GeoDataFrame) – GeoDataFrame of points from where the trips will terminate

  • rowwise (bool) – if False (default), it will calculate the cost from each origins to each destination. If true, it will calculate the cost from origin 1 to destination 1, origin 2 to destination 2 and so on.

  • destination_count (int | None) – number of closest destinations to keep for each origin. If None (default), all trips will be included. The number of destinations might be higher than the destination_count if trips have equal cost.

  • cutoff (int | float | None) – the maximum cost (weight) for the trips. Defaults to None, meaning all rows will be included. NaNs will also be removed if cutoff is specified.

  • lines (bool) – if True, returns a geometry column with straight lines between origin and destination. Defaults to False.

Return type:

DataFrame | GeoDataFrame

Returns:

A DataFrame with the weight column and the columns ‘origin’ and ‘destination’, containing the indices of the origins and destinations GeoDataFrames. If lines is True, also returns a geometry column with straight lines between origin and destination.

Examples:

Create the NetworkAnalysis instance.

>>> import sgis as sg
>>> import numpy as np
>>> np.random.seed(0)
>>> roads = sg.read_parquet_url("https://media.githubusercontent.com/media/statisticsnorway/ssb-sgis/main/tests/testdata/roads_oslo_2022.parquet")
>>> directed_roads = sg.get_connected_components(roads).loc[lambda x: x["connected"] == 1].pipe(sg.make_directed_network_norway, dropnegative=True)
>>> rules = sg.NetworkAnalysisRules(weight="minutes", directed=True)
>>> nwa = sg.NetworkAnalysis(network=directed_roads, rules=rules, detailed_log=False)

Create some origin and destination points.

>>> points = sg.read_parquet_url("https://media.githubusercontent.com/media/statisticsnorway/ssb-sgis/main/tests/testdata/points_oslo.parquet")
>>> origins = points.loc[:9, ["geometry"]]
>>> print(origins.to_string(max_rows=10, max_colwidth=50, show_dimensions=True))
                     geometry
0  POINT (263122.7 6651184.9)
1  POINT (272456.1 6653369.5)
2  POINT (270082.3 6653032.7)
3  POINT (259804.8 6650339.7)
4  POINT (272876.2 6652889.1)
5  POINT (257076.8 6652598.4)
6  POINT (268426.2 6649159.6)
7    POINT (261664.6 6650990)
8  POINT (258505.6 6654539.4)
9      POINT (259736 6650048)

[10 rows x 1 columns]
>>> destinations = points.loc[10:19, ["geometry"]]
>>> print(destinations.to_string(max_rows=10, max_colwidth=50, show_dimensions=True))
                      geometry
10  POINT (265479.7 6652146.1)
11  POINT (267812.3 6646693.7)
12    POINT (265624 6646283.7)
13    POINT (258666.6 6648408)
14  POINT (266136.8 6641975.4)
15  POINT (261829.1 6653277.7)
16  POINT (266549.9 6652070.9)
17  POINT (257867.8 6651037.6)
18  POINT (257347.7 6656069.5)
19  POINT (267030.2 6642630.6)

[10 rows x 1 columns]

Travel time from ten origins to ten destinations.

>>> od = nwa.od_cost_matrix(origins, destinations)
>>> print(od.to_string(max_rows=10, max_colwidth=50, show_dimensions=True))
    origin  destination    minutes
0        0           10   5.614867
1        1           10  12.098424
2        2           10  10.776401
3        3           10  12.359597
4        4           10  13.736730
..     ...          ...        ...
95       5           19  26.021529
96       6           19  16.017797
97       7           19  16.982346
98       8           19  24.561074
99       9           19  16.844023

[100 rows x 3 columns]

Assign aggregated values onto the origins (or destinations).

>>> origins["minutes_min"] = od.groupby("origin")["minutes"].min()
>>> origins["minutes_mean"] = od.groupby("origin")["minutes"].mean()
>>> origins["n_missing"] = len(destinations) - od.groupby("origin")["minutes"].count()
>>> print(origins.to_string(max_rows=10, max_colwidth=50, show_dimensions=True))
                     geometry  minutes_min  minutes_mean  n_missing
0  POINT (263122.7 6651184.9)     5.614867     11.711857          0
1  POINT (272456.1 6653369.5)     8.338750     17.323004          0
2  POINT (270082.3 6653032.7)     7.423471     16.285906          0
3  POINT (259804.8 6650339.7)     4.983162     12.202607          0
4  POINT (272876.2 6652889.1)     9.977056     18.961310          0
5  POINT (257076.8 6652598.4)    10.316722     17.903136          0
6  POINT (268426.2 6649159.6)     7.239519     15.471286          0
7    POINT (261664.6 6650990)     5.256121     12.346193          0
8  POINT (258505.6 6654539.4)     4.992840     15.739859          0
9      POINT (259736 6650048)     4.808400     12.308181          0

[10 rows x 4 columns]

Join the results onto the ‘origins’ via the index.

>>> joined = origins.join(od.set_index("origin"))
>>> print(joined.to_string(max_rows=10, max_colwidth=50, show_dimensions=True))
                      geometry  minutes_min  minutes_mean  n_missing  destination    minutes
0   POINT (263122.7 6651184.9)     5.614867     11.711857          0           10   5.614867
0   POINT (263122.7 6651184.9)     5.614867     11.711857          0           11  13.491910
0   POINT (263122.7 6651184.9)     5.614867     11.711857          0           12  11.642500
0   POINT (263122.7 6651184.9)     5.614867     11.711857          0           13  12.996674
0   POINT (263122.7 6651184.9)     5.614867     11.711857          0           14  14.862544
..                         ...          ...           ...        ...          ...        ...
9       POINT (259736 6650048)     4.808400     12.308181          0           15   7.935885
9       POINT (259736 6650048)     4.808400     12.308181          0           16  13.470447
9       POINT (259736 6650048)     4.808400     12.308181          0           17   4.808400
9       POINT (259736 6650048)     4.808400     12.308181          0           18  17.019517
9       POINT (259736 6650048)     4.808400     12.308181          0           19  16.844023

[100 rows x 6 columns]

Keep only travel times of 10 minutes or less. This is the same as using the cutoff parameter.

>>> ten_min_or_less = od.loc[od.minutes <= 10]
>>> joined = origins.join(ten_min_or_less.set_index("origin"))
>>> print(joined.to_string(max_rows=10, max_colwidth=50, show_dimensions=True))
                      geometry  minutes_min  minutes_mean  n_missing  destination   minutes
0   POINT (263122.7 6651184.9)     5.614867     11.711857          0         10.0  5.614867
0   POINT (263122.7 6651184.9)     5.614867     11.711857          0         15.0  5.913632
0   POINT (263122.7 6651184.9)     5.614867     11.711857          0         16.0  6.199637
1   POINT (272456.1 6653369.5)     8.338750     17.323004          0         16.0  8.338750
2   POINT (270082.3 6653032.7)     7.423471     16.285906          0         16.0  7.423471
..                         ...          ...           ...        ...          ...       ...
8   POINT (258505.6 6654539.4)     4.992840     15.739859          0         15.0  7.980121
8   POINT (258505.6 6654539.4)     4.992840     15.739859          0         18.0  4.992840
9       POINT (259736 6650048)     4.808400     12.308181          0         13.0  5.660898
9       POINT (259736 6650048)     4.808400     12.308181          0         15.0  7.935885
9       POINT (259736 6650048)     4.808400     12.308181          0         17.0  4.808400

[20 rows x 6 columns]

Keep the three fastest times from each origin. This is the same as using the destination_count parameter.

>>> three_fastest = od.loc[od.groupby("origin")["minutes"].rank() <= 3]
>>> joined = origins.join(three_fastest.set_index("origin"))
>>> print(joined.to_string(max_rows=10, max_colwidth=50, show_dimensions=True))
                      geometry  minutes_min  minutes_mean  n_missing  destination    minutes
0   POINT (263122.7 6651184.9)     5.614867     11.711857          0           10   5.614867
0   POINT (263122.7 6651184.9)     5.614867     11.711857          0           15   5.913632
0   POINT (263122.7 6651184.9)     5.614867     11.711857          0           16   6.199637
1   POINT (272456.1 6653369.5)     8.338750     17.323004          0           10  12.098424
1   POINT (272456.1 6653369.5)     8.338750     17.323004          0           11  13.974151
..                         ...          ...           ...        ...          ...        ...
8   POINT (258505.6 6654539.4)     4.992840     15.739859          0           17  10.804265
8   POINT (258505.6 6654539.4)     4.992840     15.739859          0           18   4.992840
9       POINT (259736 6650048)     4.808400     12.308181          0           13   5.660898
9       POINT (259736 6650048)     4.808400     12.308181          0           15   7.935885
9       POINT (259736 6650048)     4.808400     12.308181          0           17   4.808400

[30 rows x 6 columns]

Use set_index to use column as identifier insted of the index.

>>> origins["areacode"] = np.random.choice(["0301", "3401"], len(origins))
>>> od = nwa.od_cost_matrix(
...    origins.set_index("areacode"),
...    destinations
... )
>>> print(od.to_string(max_rows=10, max_colwidth=50, show_dimensions=True))
   origin  destination    minutes
0    0301           10   5.614867
1    3401           10  12.098424
2    3401           10  10.776401
3    0301           10  12.359597
4    3401           10  13.736730
..    ...          ...        ...
95   3401           19  26.021529
96   3401           19  16.017797
97   3401           19  16.982346
98   3401           19  24.561074
99   3401           19  16.844023

[100 rows x 3 columns]

Pair ten origins with ten destinations rowwise.

>>> rowwise_points = points.iloc[:10]
>>> points_reversed = rowwise_points.iloc[::-1]
>>> od = nwa.od_cost_matrix(rowwise_points, points_reversed, rowwise=True)
>>> print(od.to_string(max_rows=10, max_colwidth=50, show_dimensions=True))
   origin  destination    minutes
0       0            9   8.275748
1       1            8  20.526351
2       2            7  14.464923
3       3            6  17.637484
4       4            5  24.663967
5       5            4  25.611493
6       6            3  17.687899
7       7            2  14.507884
8       8            1  21.634552
9       9            0   8.189336

[10 rows x 3 columns]
precice_service_area(origins, breaks, *, dissolve=True)[source]

Precice, but slow version of the service_area method.

It finds all the network lines that can be reached within each break. Lines that are partly within the break will be split at the point where the weight value is exactly correct. Note that this takes more time than the regular ‘service_area’ method.

Parameters:
  • origins (GeoDataFrame) – GeoDataFrame of points from where the service areas will originate

  • breaks (int | float | tuple[int | float]) – one or more integers or floats which will be the maximum weight for the service areas. Calculates multiple areas for each origins if multiple breaks.

  • dissolve (bool) – If True (default), each service area will be dissolved into one long multilinestring. If False, the individual line segments will be returned.

Return type:

GeoDataFrame

Returns:

A GeoDataFrame with one row per break per origin, with a dissolved line geometry. If dissolve is False, it will return all the columns of the network.gdf as well.

See also

service_area: Faster method where lines are not cut to get precice results.

Examples:

Create the NetworkAnalysis instance.

>>> import sgis as sg
>>> from shapely.geometry import LineString
>>> roads = sg.to_gdf([
...     LineString([(0, 0), (1000, 0)]),
...     LineString([(1000, 0), (2000, 0)]),
...     LineString([(2000, 0), (3000, 0)]),
... ], crs=25833)
>>> roads["minutes"] = 5.0
>>> rules = sg.NetworkAnalysisRules(weight="minutes", directed=False)
>>> nwa = sg.NetworkAnalysis(network=roads, rules=rules)

10 minute service area for one origin point.

>>> points = sg.to_gdf([(0, 0), (500, 0), (1000, 0)], crs=25833)
>>> sa = nwa.precice_service_area(
...         points.iloc[[0]],
...         breaks=10,
...     )
>>> sa
   minutes                                           geometry
0       10  MULTILINESTRING ((1000 0, 1005 0), (1005 0, 19...

Service areas of 5, 10 and 15 minutes from two origin points.

>>> sa = nwa.precice_service_area(
...         points.iloc[:2],
...         breaks=[5, 10, 15],
...     )
>>> sa
   minutes                                           geometry
0        5  MULTILINESTRING ((0 0, 2.5 0), (2.5 0, 5 0), (...
1       10  MULTILINESTRING ((1000 0, 1005 0), (1005 0, 19...
2       15  MULTILINESTRING ((1000 0, 1005 0), (1005 0, 10...
3        5  MULTILINESTRING ((0 0, 2.5 0), (2.5 0, 5 0), (...
4       10  MULTILINESTRING ((1000 0, 1002.5 0), (1002.5 0...
5       15  MULTILINESTRING ((1000 0, 1007.5 0), (1007.5 0...
service_area(origins, breaks, *, dissolve=True)[source]

Returns the lines that can be reached within breaks (weight values).

It finds all the network lines that can be reached within each break. Lines that are only partly within the break will not be included. The index of the origins is used as values in the ‘origins’ column.

Parameters:
  • origins (GeoDataFrame) – GeoDataFrame of points from where the service areas will originate

  • breaks (int | float | tuple[int | float]) – one or more integers or floats which will be the maximum weight for the service areas. Calculates multiple areas for each origins if multiple breaks.

  • dissolve (bool) – If True (default), each service area will be dissolved into one long multilinestring. If False, the individual line segments will be returned.

Return type:

GeoDataFrame

Returns:

A GeoDataFrame with one row per break per origin, with the origin index and a dissolved line geometry. If dissolve is False, it will return each line that is part of the service area.

See also

precice_service_area: Equivelent method where lines are also cut to get precice results.

Examples:

Create the NetworkAnalysis instance.

>>> import sgis as sg
>>> from shapely.geometry import LineString
>>> roads = sg.to_gdf([
...     LineString([(0, 0), (1000, 0)]),
...     LineString([(1000, 0), (2000, 0)]),
...     LineString([(2000, 0), (3000, 0)]),
... ], crs=25833)
>>> roads["minutes"] = 5.0
>>> rules = sg.NetworkAnalysisRules(weight="minutes", directed=False)
>>> nwa = sg.NetworkAnalysis(network=roads, rules=rules)

10 minute service area for three origin points.

>>> points = sg.to_gdf([(0, 0), (500, 0), (1000, 0)], crs=25833)
>>> service_areas = nwa.service_area(
...         points.loc[:2],
...         breaks=10,
... )
>>> print(service_areas.to_string(max_rows=10, max_colwidth=50, show_dimensions=True))
   origin  minutes                                           geometry
0       0       10  MULTILINESTRING ((0 0, 500 0), (500 0, 1000 0)...
1       1       10  MULTILINESTRING ((0 0, 500 0), (500 0, 1000 0)...
2       2       10  MULTILINESTRING ((2000 0, 3000 0), (0 0, 500 0...

[3 rows x 3 columns]

Service areas of 5, 10 and 15 minutes from two origin points.

>>> service_areas = nwa.service_area(
...         points.iloc[:2],
...         breaks=[5, 10, 15],
... )
>>> print(service_areas.to_string(max_rows=10, max_colwidth=50, show_dimensions=True))
   origin  minutes                                         geometry
0       0        5  MULTILINESTRING ((0 0, 500 0), (500 0, 1000 0))
1       0       10                      LINESTRING (1000 0, 2000 0)
2       0       15                      LINESTRING (2000 0, 3000 0)
3       1        5  MULTILINESTRING ((0 0, 500 0), (500 0, 1000 0))
4       1       10                      LINESTRING (1000 0, 2000 0)
5       1       15                      LINESTRING (2000 0, 3000 0)

[6 rows x 3 columns]